[ovs-discuss] Tuple Space Search implementation in the MegaFlow Cache

[Ben Pfaff]

On Mon, Oct 24, 2016 at 04:55:52PM -0400, thibaut stimpfling wrote:
> I am looking for implementation details of the Tuple Space Search
> algorithm in the MegaFlow Cache. More specifically, I am interested in
> how ranges are handled by the tuple space search lookup algorithm. Is
> each range converted into (a) prefix(es) ?

Yes.

In more detail, OVS itself doesn't handle ranges at all.  Instead,
whatever feeds flows to it must break ranges into multiple flows that
use prefix matching.

> Are ranges converted into Nesting IDs/Nesting Levels, as described in
> the paper "Packet classification using tuple space search” ? 

No.

I've read that paper a couple of times and I've always failed to
understand that part of it.  I'd very much appreciate it if someone
would explain it to me.

> If so, which data structure are you using to find the nesting Id /
> nesting level associated with a packet header field ?
> 
> Can you point me to any documentation that could answer this question ?

Did you read our NSDI paper?  It's all about the OVS classifier.
https://www.usenix.org/node/188961